Environmental Pollution 188 (2014) 109e117
Contents lists available at ScienceDirect
Environmental Pollution journal homepage: www.elsevier.com/locate/envpol
Partitioning of magnetic particles in PM10, PM2.5 and PM1 aerosols in the urban atmosphere of Barcelona (Spain) María Aránzazu Revuelta a, Gregg McIntosh b, c, *, Jorge Pey d, e, Noemi Pérez d, Xavier Querol d, Andrés Alastuey d a
State Meteorological Agency of Spain (AEMET), c/Leonardo Prieto Castro 8, 28071 Madrid, Spain Geosciences Institute (IGEO, UCM-CSIC), Geology Faculty, c/ Jose Antonio Navais 2, Ciudad Universitaria, 28040 Madrid, Spain Department of Earth Physics, Astronomy and Astrophysics I, Physics Faculty, Complutense University of Madrid, Avda Complutense s/n, Ciudad Universitaria, 28040 Madrid, Spain d Institute of Environmental Assessment and Water Research, Spanish Research Council (IDAEA-CSIC), c/Jordi Girona 18-26, 08034 Barcelona, Spain e Currently at Aix-Marseille Université, Laboratoire Chimie Environnement, France b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 October 2013 Received in revised form 24 January 2014 Accepted 27 January 2014
A combined magnetic-chemical study of 15 daily, simultaneous PM10ePM2.5ePM1 urban background aerosol samples has been carried out. The magnetic properties are dominated by non-stoichiometric magnetite, with highest concentrations seen in PM10. Low temperature magnetic analyses showed that the superparamagnetic fraction is more abundant when coarse, multidomain particles are present, confirming that they may occur as an oxidized outer shell around coarser grains. A strong association of the magnetic parameters with a vehicular PM10 source has been identified. Strong correlations found with Cu and Sb suggests that this association is related to brake abrasion emissions rather than exhaust emissions. For PM1 the magnetic remanence parameters are more strongly associated with crustal sources. Two crustal sources are identified in PM1, one of which is of North African origin. The magnetic particles are related to this source and so may be used to distinguish North African dust from other sources in PM1. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Environmental magnetism Atmospheric aerosol Source apportionment
1. Introduction Particulate matter (PM) in the atmosphere is a major environmental concern, especially with respect to its impact on human health. Airborne PM can be defined by its aerodynamic properties. PM10 (the PM fraction with aerodynamic diameter 0.5) loadings on Fe are seen in all three PM sizes, although the loading decreases with decreasing PM size. The contribution of this component is highest in LOC-dominated samples, implying that road traffic is the most important local anthropogenic PM source. TC and NO3 are typical tracers of combustion emissions, whereas Sb is commonly attributed to abrasion emissions (Hildemann et al., 1991; Garg et al., 2000; Amato et al., 2009; Pey et al., 2010b). k has a high loading on all three PM sizes, whilst kARM and Mrs loadings decrease with decreasing PM size. A single crustal component has been identified in PM10 and PM2.5 by high loadings on Al2O3, Ca, K, Mg, Fe, Li, Ti, Ga, Rb, Sr, La and Ce. Its contribution is highest in the NAF-dominated sample. In PM1 the crustal contribution is split into two components; crustal-1 having high loadings on Al2O3, Ca, Na, Mg, Ti, La and Ce (typically associated with silicate- and carbonate-rich dusts) and crustal-2 having high loadings on K, Fe, Li, Ti, Mn, Cu, Zn, Ga, Rb and Sr (typically associated with clay-rich dusts). Crustal-1 shows its highest contribution in the W ATL samples and crustal-2 dominates the NAF samples. kARM and Mrs have high loadings on the PM2.5 crustal components and on the PM1 crustal-2 component. k does not show salient loadings on any of the crustal components. The component structures and compositions were stable with respect to the “leave on out” tests, with mean loadings that were in close agreement with the original 15 sample model (see Supplementary material). The sample with the strongest influence was the NAF-dominated sample, which impacted on the crustal and vehicular components in all three PM sizes. Removing the NAFdominated sample had two effects on the PM10 and PM2.5 fractions; (a) reducing the relative importance of the crustal
114
M.A. Revuelta et al. / Environmental Pollution 188 (2014) 109e117
Fig. 3. (aec) Representative hysteresis (blue) and back-field IRM (red) curves, with B ¼ applied field and M ¼ magnetization. (d) Day plot of magnetization ratios (Mrs/Ms) and coercivity ratios (Bcr/Bc). The SSD-MD and SSD-SP magnetite unmixing lines of Dunlop (2002) are shown. For definitions of synaptic wind conditions see Fig. 1 caption. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
component and reducing the loadings on kARM, Mrs, Fe, Zn and Ba, and (b) increasing the relative importance of the vehicular component and increasing the loadings on kARM, Mrs and Fe so that they loaded uniquely on the vehicular component. For PM1, removing the NAF-dominated sample effectively lead to the loss of the crustal-2 component and to higher and unique loadings on kARM, Mrs, Fe, Mn and Cu on the vehicular component. Including the NAF-dominated sample leads to the definition of a discrete crustal source in PM1, which is not recognized when considering only those samples with mixed synoptic conditions (LOC þ NAF, NAF þ EW), and which leads to a better discrimination of the potential PM sources. However, the magnetic results are strongly affected by the removal of the NAF-dominated sample. This can be explained by postulating two sources of strongly
Fig. 4. Thermomagnetic curves, where T ¼ temperature and M ¼ magnetization. Heating (cooling) branches shown in red (blue). For definitions of synaptic wind conditions see Fig. 1 caption. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
magnetic PM; a vehicular source with high loadings on k, kARM and Mrs and a NAF crustal source with high loadings on kARM and Mrs. This is supported by the magnetic hysteresis data, which show clear differences between the LOC and NAF samples (Fig. 3d). Reducing the importance of the NAF crustal source leads to the association of k, kARM and Mrs solely with the vehicular source, whereas increasing its importance increases the loadings of kARM and Mrs on the NAF crustal component. k is unaffected by the changing source importance, which suggests that it is solely influenced by the vehicular source. The association of k with the vehicular source may be explained by the presence of sub-micron, SP particles that are vehicular in
Fig. 5. Thermal demagnetization of IRM acquired at 268 C, where T ¼ temperature and IRM268/IRM ¼ normalized IRM. For definitions of synaptic wind conditions see Fig. 1 caption.
M.A. Revuelta et al. / Environmental Pollution 188 (2014) 109e117
Fig. 6. Biplot of susceptibility of anhysteretic remanence, kARM, versus saturation isothermal remanence, Mrs.
origin rather than crustal. This would act to bias k towards the vehicular source component more strongly than either kARM or Mrs, which can be observed in the markedly higher loadings on k (Table 2). The loadings on k of the vehicular component are lower for PM2.5 and PM1. This may be due to the reduced relative importance of SP particles in the finer PM sizes (Section 3.2.1) and to the larger relative errors associated with the finer PM sizes (mean error values increase from 2% in PM10 to 17% in PM1). The strongest association of kARM and Mrs with the vehicular component is seen in PM10, which indicates that the coarsest magnetic particles (>2.5 mm) are predominantly vehicular in origin, confirming the results of previous studies where industrial sources are expected to play a lesser role than traffic (e.g. Sagnotti et al., 2006; Saragnese et al., 2011). kARM and Mrs exhibit decreasing loadings on the vehicular factor as PM size decreases (Table 2). Both parameters are more precisely determined than k (mean errors
Contents lists available at ScienceDirect
Environmental Pollution journal homepage: www.elsevier.com/locate/envpol
Partitioning of magnetic particles in PM10, PM2.5 and PM1 aerosols in the urban atmosphere of Barcelona (Spain) María Aránzazu Revuelta a, Gregg McIntosh b, c, *, Jorge Pey d, e, Noemi Pérez d, Xavier Querol d, Andrés Alastuey d a
State Meteorological Agency of Spain (AEMET), c/Leonardo Prieto Castro 8, 28071 Madrid, Spain Geosciences Institute (IGEO, UCM-CSIC), Geology Faculty, c/ Jose Antonio Navais 2, Ciudad Universitaria, 28040 Madrid, Spain Department of Earth Physics, Astronomy and Astrophysics I, Physics Faculty, Complutense University of Madrid, Avda Complutense s/n, Ciudad Universitaria, 28040 Madrid, Spain d Institute of Environmental Assessment and Water Research, Spanish Research Council (IDAEA-CSIC), c/Jordi Girona 18-26, 08034 Barcelona, Spain e Currently at Aix-Marseille Université, Laboratoire Chimie Environnement, France b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 October 2013 Received in revised form 24 January 2014 Accepted 27 January 2014
A combined magnetic-chemical study of 15 daily, simultaneous PM10ePM2.5ePM1 urban background aerosol samples has been carried out. The magnetic properties are dominated by non-stoichiometric magnetite, with highest concentrations seen in PM10. Low temperature magnetic analyses showed that the superparamagnetic fraction is more abundant when coarse, multidomain particles are present, confirming that they may occur as an oxidized outer shell around coarser grains. A strong association of the magnetic parameters with a vehicular PM10 source has been identified. Strong correlations found with Cu and Sb suggests that this association is related to brake abrasion emissions rather than exhaust emissions. For PM1 the magnetic remanence parameters are more strongly associated with crustal sources. Two crustal sources are identified in PM1, one of which is of North African origin. The magnetic particles are related to this source and so may be used to distinguish North African dust from other sources in PM1. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Environmental magnetism Atmospheric aerosol Source apportionment
1. Introduction Particulate matter (PM) in the atmosphere is a major environmental concern, especially with respect to its impact on human health. Airborne PM can be defined by its aerodynamic properties. PM10 (the PM fraction with aerodynamic diameter 0.5) loadings on Fe are seen in all three PM sizes, although the loading decreases with decreasing PM size. The contribution of this component is highest in LOC-dominated samples, implying that road traffic is the most important local anthropogenic PM source. TC and NO3 are typical tracers of combustion emissions, whereas Sb is commonly attributed to abrasion emissions (Hildemann et al., 1991; Garg et al., 2000; Amato et al., 2009; Pey et al., 2010b). k has a high loading on all three PM sizes, whilst kARM and Mrs loadings decrease with decreasing PM size. A single crustal component has been identified in PM10 and PM2.5 by high loadings on Al2O3, Ca, K, Mg, Fe, Li, Ti, Ga, Rb, Sr, La and Ce. Its contribution is highest in the NAF-dominated sample. In PM1 the crustal contribution is split into two components; crustal-1 having high loadings on Al2O3, Ca, Na, Mg, Ti, La and Ce (typically associated with silicate- and carbonate-rich dusts) and crustal-2 having high loadings on K, Fe, Li, Ti, Mn, Cu, Zn, Ga, Rb and Sr (typically associated with clay-rich dusts). Crustal-1 shows its highest contribution in the W ATL samples and crustal-2 dominates the NAF samples. kARM and Mrs have high loadings on the PM2.5 crustal components and on the PM1 crustal-2 component. k does not show salient loadings on any of the crustal components. The component structures and compositions were stable with respect to the “leave on out” tests, with mean loadings that were in close agreement with the original 15 sample model (see Supplementary material). The sample with the strongest influence was the NAF-dominated sample, which impacted on the crustal and vehicular components in all three PM sizes. Removing the NAFdominated sample had two effects on the PM10 and PM2.5 fractions; (a) reducing the relative importance of the crustal
114
M.A. Revuelta et al. / Environmental Pollution 188 (2014) 109e117
Fig. 3. (aec) Representative hysteresis (blue) and back-field IRM (red) curves, with B ¼ applied field and M ¼ magnetization. (d) Day plot of magnetization ratios (Mrs/Ms) and coercivity ratios (Bcr/Bc). The SSD-MD and SSD-SP magnetite unmixing lines of Dunlop (2002) are shown. For definitions of synaptic wind conditions see Fig. 1 caption. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
component and reducing the loadings on kARM, Mrs, Fe, Zn and Ba, and (b) increasing the relative importance of the vehicular component and increasing the loadings on kARM, Mrs and Fe so that they loaded uniquely on the vehicular component. For PM1, removing the NAF-dominated sample effectively lead to the loss of the crustal-2 component and to higher and unique loadings on kARM, Mrs, Fe, Mn and Cu on the vehicular component. Including the NAF-dominated sample leads to the definition of a discrete crustal source in PM1, which is not recognized when considering only those samples with mixed synoptic conditions (LOC þ NAF, NAF þ EW), and which leads to a better discrimination of the potential PM sources. However, the magnetic results are strongly affected by the removal of the NAF-dominated sample. This can be explained by postulating two sources of strongly
Fig. 4. Thermomagnetic curves, where T ¼ temperature and M ¼ magnetization. Heating (cooling) branches shown in red (blue). For definitions of synaptic wind conditions see Fig. 1 caption. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
magnetic PM; a vehicular source with high loadings on k, kARM and Mrs and a NAF crustal source with high loadings on kARM and Mrs. This is supported by the magnetic hysteresis data, which show clear differences between the LOC and NAF samples (Fig. 3d). Reducing the importance of the NAF crustal source leads to the association of k, kARM and Mrs solely with the vehicular source, whereas increasing its importance increases the loadings of kARM and Mrs on the NAF crustal component. k is unaffected by the changing source importance, which suggests that it is solely influenced by the vehicular source. The association of k with the vehicular source may be explained by the presence of sub-micron, SP particles that are vehicular in
Fig. 5. Thermal demagnetization of IRM acquired at 268 C, where T ¼ temperature and IRM268/IRM ¼ normalized IRM. For definitions of synaptic wind conditions see Fig. 1 caption.
M.A. Revuelta et al. / Environmental Pollution 188 (2014) 109e117
Fig. 6. Biplot of susceptibility of anhysteretic remanence, kARM, versus saturation isothermal remanence, Mrs.
origin rather than crustal. This would act to bias k towards the vehicular source component more strongly than either kARM or Mrs, which can be observed in the markedly higher loadings on k (Table 2). The loadings on k of the vehicular component are lower for PM2.5 and PM1. This may be due to the reduced relative importance of SP particles in the finer PM sizes (Section 3.2.1) and to the larger relative errors associated with the finer PM sizes (mean error values increase from 2% in PM10 to 17% in PM1). The strongest association of kARM and Mrs with the vehicular component is seen in PM10, which indicates that the coarsest magnetic particles (>2.5 mm) are predominantly vehicular in origin, confirming the results of previous studies where industrial sources are expected to play a lesser role than traffic (e.g. Sagnotti et al., 2006; Saragnese et al., 2011). kARM and Mrs exhibit decreasing loadings on the vehicular factor as PM size decreases (Table 2). Both parameters are more precisely determined than k (mean errors
